Asynchronous Transfer Mode (ATM) networks have become increasingly popular for both wide area networks and local area networks. In an ATM network all information is transported through the network in relatively short blocks called cells. Information flow through the network is along paths called virtual channels which are set up using a series of tables implemented in switching nodes that comprise the network. Cells on a particular virtual channel follow the same path through the network and are delivered to the destination in the order in which they are received.
When communications are initiated in an ATM network a request is made for a connection. As part of the connection request, the quality of service (QoS) of the request is specified by the requestor. Quality of service characteristics which may be specified in ATM include cell transfer delay (network latency), cell delay variation (jitter), cell transfer capacity (average and peak allowed rates), cell error ratio, cell loss ratio and cell misinsertion ratio. These QoS parameters may be used by the ATM nodes to determine the route of the connection and in determining allowable node utilization.
Routing in an ATM network is performed by an ATM node which attempts to find a feasible route for a virtual connection from a source to a destination. An ATM connection is not set up unless and until the network determines that there is sufficient capacity for the new connection. This determination is based upon the characteristics of the network, the existing traffic on the network and the requested QoS for the connection. If the requested QoS cannot be provided then the requested connection is not accepted. The function in ATM which determines whether a connection request is accepted is referred to as call admission control (CAC).
In ATM, the CAC function is typically carried out at two places; the entry point into the ATM network and at each node in the connection path. At the entry point into the ATM network the CAC function determines a feasible route for the connection request based on the QoS requested, and either accepts or rejects the request based on this determination. To perform the entry point CAC function, the entry point should have available information about the current utilization of the other nodes in the ATM network. This information may then be utilized to determine if the nodes along the selected route can satisfy the QoS requirements of the requested connection.
At each node in a connection route, a local CAC function determines if the node will accept the connection request. This CAC function utilizes the requested QoS and information about the existing connections through the node to determine if the requested QoS level may be achieved. If so, the connection request is accepted.
As is apparent from the above discussion, the call acceptance procedure of an ATM system may impact on the performance of the network. If too many connection requests are accepted then the QoS of the connections may be degraded and additional, possibly costly, resources may be required to handle the network's traffic. If too few connection requests are accepted, then the network will be under-utilized. The call admission problem is further complicated by the fact that the call admission procedure is typically carried out in real-time as call requests are received. The limits placed on the complexity of the call admission procedure by the real-time requirement may potentially result in compromises resulting in either too conservative a call admission procedure or too optimistic a procedure. Finally, the call admission problem is also limited by the information available to determine the characteristics of traffic through the network.
If the network makes a decision to admit a new connection, resources such as link bandwidth and buffers must be reserved to guarantee that the connection receives its guaranteed QoS. For certain types of traffic with well-known characteristics (such as voice traffic) it is straightforward to determine the amount of bandwidth needed to guarantee QoS. Where traffic is more variable (bursty), however, the network must make a more difficult decision regarding the appropriate level of bandwidth resources to set aside for the new connection. The equivalent bandwidth of a connection is defined as an "average" (or steady-state) amount of bandwidth needed to be reserved to carry the traffic of that connection when it is sharing link resources with other connections that have been similarly accommodated. CAC mechanisms based on equivalent bandwidth are typically simple in that the determination of whether a given set of connections can be accommodated without violating their QoS requirements reduces to comparing the sum of the equivalent bandwidths of individual sources to the link capacity.
Although generally simple to implement, equivalent bandwidth CAC functions are highly conservative when the buffer size is small or moderate. Thus, utilization of an equivalent bandwidth approach to call admission may result in fewer connection requests being accepted than could be accommodated by an ATM network or link in an ATM network. Accordingly, additional resources in the network may be required to handle the network traffic than would otherwise be needed if the network resources were more efficiently utilized.
A second approach to call admission control based on bandwidth is the Gaussian approximation based on a zero-buffer assumption. If the number of sources being multiplexed (N) is sufficiently large, the aggregate traffic can be approximated by a Gaussian process with mean rate ##EQU1##
and variance ##EQU2##
While call admission based on a Gaussian process may provide increased efficiency over an equivalent bandwidth approach, in a system having a buffer, the buffer's capacity to absorb traffic bursts is ignored, thus, resulting in under-utilization of the network. Furthermore, when N is small, the Gaussian approximation will not be valid which may result in extremely conservative bandwidth determinations.
Various other hybrid systems have also been proposed, however, these systems may also have limitations. For example, a system utilizing a highly non-linear function of the individual equivalent bandwidths to determine the admissibility of a given set of traffic sources may be over-optimistic in certain situations, thus, resulting in higher cell loss ratios than the specified QoS of the connections.
In light of the above discussion, a need exists for improvements in the mechanisms for accepting connections in ATM and other networks.